Report

Introduction to Geographic Information Systems at TU Dresden

Author

Tuan Linh Tran, Andrea Češková, Tobias Gruner

1 Goal of the analysis

Research question: How does the relationship between population density and temperature vary across different land use types around Netherlands?

2 Data

2.1 Outcome variable: Temperature

Figure 1: Mean Temperature by Region

2.2 Summary Statistics

Table 1: Available valid data
Land Use Type Total Pixels Valid Population Pixels % Valid
Arable land 802 623 77.7
Artificial, non-agricultural vegetated areas 1036 978 94.4
Forests 1644 1036 63.0
Heterogeneous agricultural areas 888 728 82.0
Industrial, commercial and transport units 1109 1066 96.1
Inland waters 420 193 46.0
Inland wetlands 278 134 48.2
Marine waters 5 3 60.0
Maritime wetlands 74 29 39.2
Mine, dump and construction sites 258 226 87.6
Open spaces with little or no vegetation 58 28 48.3
Pastures 736 545 74.0
Permanent crops 75 57 76.0
Scrub and/or herbaceous vegetation associations 603 301 49.9
Urban fabric 1555 1540 99.0
Note: Calculated using population raster with NA values removed.

Although some land use categories such as Inland waters, Wetlands, and Sparse vegetation have a low proportion of valid population pixels (often below 50%), we chose to retain them in the Table 2 table for the sake of completeness and visual consistency.

However, we acknowledge that the population means and correlation values in these categories may be statistically unreliable due to sparse or uneven data coverage. In particular, categories like Marine waters include only a few valid observations, resulting in potentially misleading or exaggerated correlation coefficients. We document this in the Table 1.

To address this, we highlighted such rows in the Table 2 and excluded their population and correlation values from the Figure 2, to prevent visual misinterpretation while preserving transparency in reporting.

Table 2: Summary statistics
Land Use Type Sample Size Mean Temperature (°C) SD Temperature (°C) Mean Population Median Population Correlation temperature ~ population
Forests 1,644 11.16 0.29 2.57 1.49 0.08
Urban fabric 1,555 11.21 0.27 7.68 5.91 0.11
Industrial, commercial and transport units 1,109 11.22 0.24 6.50 4.65 0.02
Artificial, non-agricultural vegetated areas 1,036 11.19 0.26 7.71 4.92 0.03
Heterogeneous agricultural areas 888 11.23 0.27 2.82 1.87 0.14
Arable land 802 11.23 0.26 2.88 1.81 0.10
Pastures 736 11.22 0.27 3.57 2.19 0.05
Scrub and/or herbaceous vegetation associations 603 11.15 0.31 2.16 1.15 0.11
Inland waters 420 11.17 0.28 3.72 2.09 0.18
Inland wetlands 278 11.15 0.28 2.46 1.19 0.05
Mine, dump and construction sites 258 11.22 0.25 5.07 2.74 0.00
Permanent crops 75 11.31 0.14 2.21 1.63 -0.13
Maritime wetlands 74 11.37 0.26 1.12 0.57 0.30
Open spaces with little or no vegetation 58 11.26 0.24 2.13 0.81 0.20
Marine waters 5 11.17 0.35 2.47 2.62 -0.69
Source: Data from Netherlands 2019: Temperature (CHELSA), Population (WorldPop), Land Use (Corine)
Figure 2: Correlation between Temperature and Population by Land Use Type across the Netherlands

3 Regression models

Table 3: Regression Results
Outcome Variable: Temperature in °C
Variable Model 1: Population Only1 Model 2: Land Use Only2 Model 3: Full Model3
Intercept 11.158*** (0.007) 11.206*** (0.004) 11.154*** (0.012)
Log(Population + 1) 0.031*** (0.004) 0.027*** (0.006)
Log(Pop) × Land Use: Agricultural areas 0.026* (0.012)
Log(Pop) × Land Use: Forest and semi natural areas 0.041** (0.014)
Log(Pop) × Land Use: Water bodies 0.076** (0.029)
Log(Pop) × Land Use: Wetlands -0.007 (0.034)
landuseAgricultural areas 0.022** (0.007) 0.009 (0.019)
landuseForest and semi natural areas -0.045*** (0.007) -0.046* (0.020)
landuseWater bodies -0.035* (0.014) -0.125** (0.043)
landuseWetlands -0.007 (0.015) 0.021 (0.040)
Standard errors in parentheses. Significance: † p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001
1 n = 7,486 | R² = 0.007 | Adj. R² = 0.007 | RMSE = 0.262
2 n = 9,528 | R² = 0.008 | Adj. R² = 0.008 | RMSE = 0.271
3 n = 7,486 | R² = 0.016 | Adj. R² = 0.014 | RMSE = 0.261

The regressions models use artificial surfaces as the baseline reference category for all land use comparisons. This means all land use coefficient should be interpreted relative to the temperature patterns observed in artificial surfaces. The analysis of temperature patterns across the Netherlands reveal that areas with higher population density consistently experience elevated temperatures. Model 1 demonstrates that a logarithmic increase in population density corresponds to a 0.031°C increase in temperature in the Model 1 and 0.027°C in the full model.

Relative to artificial surfaces, different land use types exhibit distinct temperature patterns. Agricultural areas show slightly higher temperatures (+0.009-0.022°C) compared to artificial surfaces, suggesting similar thermal characteristics between urban and agricultural environments. In contrast, forested and semi-natural areas provide significant cooling benefits, with temperatures 0.045-0.046°C lower than artificial surfaces. Water bodies also demonstrate cooling effects, particularly pronounced in the full model (-0.125°C), while wetlands show no significant temperature difference from artificial surfaces.

The interaction effects in the full model reveal how population density affects temperature differently across land use types compared to artificial surfaces. Water bodies show the strongest interaction effect (0.076°C), meaning that the population-temperature relationship is much steeper near water bodies than in built environments. Similarly, forested areas (0.041°C) and agricultural areas (0.026°C) show positive interactions, indicating that these land uses are more sensitive to population-related warming than artificial surfaces. Wetlands show no significant interaction effect, suggesting they respond similarly to population pressure as built environments.

These findings suggest that while natural land uses typically provide thermal benefits compared to artificial surfaces, they may be more vulnerable to population-related warming effects. The results highlight that the cooling advantages of forests and water bodies can be disproportionately affected by increased population density, making their preservation in low-density areas crucial for maintaining local climate benefits.

4 References

European Environment Agency. 2020. “CORINE Land Cover 2018 (Version 2020_20u1).” Copenhagen. https://land.copernicus.eu/en/products/corine-land-cover.
Hijmans, Robert J., Luigi Guarino, and Marthur Pushkar. 2023. “GADM: Database of Global Administrative Areas (Version 4.1).” https://gadm.org/.
Karger, Dirk Nikolaus, Olaf Conrad, Jürgen Böhner, Tobias Kawohl, Holger Kreft, Rodrigo Willber Soria-Auza, Niklaus E. Zimmermann, H. Peter Linder, and Michael Kessler. 2017. “Climatologies at High Resolution for the Earth’s Land Surface Areas.” Scientific Data 4 (1): 1–20. https://doi.org/10.1038/sdata.2017.122.
Tatem, Andrew J. 2017. “WorldPop, Open Data for Spatial Demography.” Scientific Data 4 (1): 1–4. https://doi.org/10.1038/sdata.2017.4.